Wan 3.0 Video API

Model Introduction

Wan 3.0 generates video from text, first-frame or first-and-last-frame images, and reference images, videos, or audio. Files and web pages can also supply source material. The same endpoint handles text-to-video, image-to-video, and reference-to-video requests.

Model	Description
wan3.0-video	Standard version
wan3.0-video-prime	Faster version with the same request format

Generation is asynchronous: create a task, save output.task_id, then query its status. Once it succeeds, retrieve the video from output.video_url.

API Overview

Item	Value
Base URL	https://api.exchangetoken.ai
Create task	POST /api/v1/services/aigc/video-generation/video-synthesis
Query task	GET /api/v1/tasks/{task_id}
Authentication	Authorization: Bearer YOUR_API_KEY
Create request format	Content-Type: application/json

Replace YOUR_API_KEY with an API key from the Exchange Token console. Use the same API key to create and query a task.

Create a Video Generation Task

POST https://api.exchangetoken.ai/api/v1/services/aigc/video-generation/video-synthesis

Request Fields

Field	Type	Required	Description
model	string	Yes	wan3.0-video or wan3.0-video-prime
input	object	Yes	Provide at least one of prompt or media
input.prompt	string	Conditional	Describe the scene, action, camera movement, or sound
input.media	object[]	Conditional	Media items, each with type and url
parameters	object	No	Generation settings, listed below

Generation Parameters

Place these fields inside the parameters object.

The values and defaults below describe Bailian's parameter contract; the upstream service validates them. Exchange Token retains model routing permissions and billing configuration checks.

Field	Type	Default	Description
resolution	string	1080P	480P, 720P, or 1080P; uppercase P
duration	integer	5	Output seconds: 2–30, or -1 for automatic duration
ratio	string	adaptive	adaptive, 21:9, 16:9, 4:3, 1:1, 3:4, or 9:16
audio	boolean	true	Generate sound
prompt_extend	boolean	true	Rewrite the prompt; required for file or link
seed	integer	-1	Random (-1) or 0–2147483647; results may vary
watermark	boolean	false	Add a watermark

Input and output video must total at most 30 seconds. For automatic duration, read the actual duration from the query response.

Media and Generation Scenarios

Scenario	type in input.media	Usage
Text-to-video	Omit media	Describe the video in input.prompt
First-frame image-to-video	first_frame	One starting image
First-and-last-frame image-to-video	first_frame + last_frame	One image per endpoint
Reference-to-video	reference_image, reference_video, reference_audio	Combine reference media
File or web page	file or link	Choose one; references allowed; enable prompt_extend

Do not mix frame inputs with references, files, or links. Use publicly accessible HTTP(S) URLs; images also accept Base64 Data URLs. For format, size, count, and duration limits, see the official media parameter reference.

Text-to-Video Example

Replace YOUR_API_KEY with your Exchange Token API key. This example requests a five-second 720P video with sound.

curl -sS -X POST "https://api.exchangetoken.ai/api/v1/services/aigc/video-generation/video-synthesis" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "wan3.0-video",
    "input": {
      "prompt": "A white sailboat glides along the coast at sunrise. The camera follows smoothly as sunlight catches the water, with gentle sounds of waves."
    },
    "parameters": {
      "resolution": "720P",
      "duration": 5,
      "ratio": "16:9",
      "audio": true
    }
  }'

To use the faster version, change model to wan3.0-video-prime.

First-and-Last-Frame Example

Replace the https://example.com/... placeholders with publicly accessible media URLs that meet the input limits. For a first-frame-only request, remove the last_frame item.

curl -sS -X POST "https://api.exchangetoken.ai/api/v1/services/aigc/video-generation/video-synthesis" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "wan3.0-video",
    "input": {
      "prompt": "The camera slowly moves closer as the flower bud opens, transitioning naturally into the full bloom shown in the last frame.",
      "media": [
        {"type": "first_frame", "url": "https://example.com/first-frame.jpg"},
        {"type": "last_frame", "url": "https://example.com/last-frame.jpg"}
      ]
    },
    "parameters": {
      "resolution": "720P",
      "duration": 5,
      "ratio": "adaptive"
    }
  }'

Reference-to-Video Example

This example supplies an image, a video, and an audio reference. The prompt uses 图1, 视频1, and 音频1 to refer to the first image, video, and audio item respectively; each media type is numbered independently.

curl -sS -X POST "https://api.exchangetoken.ai/api/v1/services/aigc/video-generation/video-synthesis" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "wan3.0-video-prime",
    "input": {
      "prompt": "Animate the character in 图1 using the movement from 视频1 and the voice from 音频1. Keep the character appearance consistent.",
      "media": [
        {"type": "reference_image", "url": "https://example.com/character.jpg"},
        {"type": "reference_video", "url": "https://example.com/motion.mp4"},
        {"type": "reference_audio", "url": "https://example.com/voice.mp3"}
      ]
    },
    "parameters": {
      "resolution": "720P",
      "duration": 5,
      "ratio": "16:9",
      "audio": true
    }
  }'

Create Response

{
  "output": {
    "task_id": "sg_vid_xxxxxxxxxxxxxxxx",
    "task_status": "PENDING"
  },
  "request_id": "request_xxxxxxxxxxxxxxxx"
}

Save output.task_id and start querying. request_id is for troubleshooting and cannot replace the task ID. Once you have a task ID, query that task: repeating the create request creates another task.

Query a Task and Retrieve the Result

Replace the ID in the path with output.task_id from the create response. Use the API key that created the task.

curl -sS "https://api.exchangetoken.ai/api/v1/tasks/sg_vid_xxxxxxxxxxxxxxxx" \
  -H "Authorization: Bearer YOUR_API_KEY"

Poll every 15 seconds and set an application waiting limit. Continue while the status is PENDING or RUNNING. If your waiting limit is reached, keep the task ID to resume querying later.

Task Statuses

output.task_status	Meaning and action
PENDING	Queued; continue polling
RUNNING	Generating; continue polling
SUCCEEDED	Complete; read output.video_url
FAILED	Failed; inspect output.code, if present, and output.message
CANCELED	Canceled; stop polling
UNKNOWN	The provider reports a missing task or unknown state; stop polling and investigate

HTTP 200 means the query succeeded. Check output.task_status to determine whether video generation succeeded.

Successful Response Example

This example includes 2.25 seconds of input reference video and 5 seconds of output video.

{
  "request_id": "request_xxxxxxxxxxxxxxxx",
  "output": {
    "task_id": "sg_vid_xxxxxxxxxxxxxxxx",
    "task_status": "SUCCEEDED",
    "video_url": "https://example.com/generated-video.mp4"
  },
  "usage": {
    "video_count": 1,
    "duration": 7.25,
    "input_video_duration": 2.25,
    "output_video_duration": 5,
    "SR": 720,
    "fps": 30,
    "ratio": "16:9"
  }
}

Responses may also include output.submit_time, output.scheduled_time, output.end_time, and output.orig_prompt. The video URL is temporary; download or copy the result to your storage promptly.

Usage and Billing

The usage object in a successful result contains actual video usage. Usage may be unavailable while a task is pending or running.

Field	Description
usage.video_count	Number of generated videos
usage.duration	Total billable seconds, including input reference video and output video; fractional values are supported
usage.input_video_duration	Input reference video seconds; 0 without video input
usage.output_video_duration	Actual output video seconds
usage.SR	Actual resolution tier, such as 480, 720, or 1080; the field name is uppercase
usage.fps	Output frame rate
usage.ratio	Output aspect ratio

Only successful tasks are billed. Querying a task adds no generation charge. Charges use the selected model, actual resolution, and usage.duration; the audio toggle does not change the unit price. Do not substitute the requested parameters.duration for actual usage or add input video seconds to usage.duration again. The example above has 7.25 billable seconds.

Failed Task Example

{
  "output": {
    "task_id": "sg_vid_xxxxxxxxxxxxxxxx",
    "task_status": "FAILED",
    "message": "Video generation failed."
  }
}

Task failure details appear inside output. An output.code field is also returned when a specific error code is available.

Request Errors

Platform Validation Errors

Errors generated by Exchange Token, such as an invalid platform API key, insufficient balance, or missing billing configuration, use this format. This example is returned with HTTP 503:

{
  "error": {
    "type": "billing_unavailable",
    "message": "media billing configuration unavailable for routed model"
  }
}

For HTTP 404, check the full task ID and confirm you are using the API key that created the task. For parameter errors, check the model name, resolution capitalization, duration, and media combination.

Upstream Task Creation Errors

The current Wan 3.0 create endpoint also wraps Bailian request errors in an error object: the upstream code becomes error.type, and message becomes error.message. For example, a rate limit error can produce:

{
  "error": {
    "type": "Throttling",
    "message": "rate limited"
  }
}

Non-2xx upstream HTTP status codes are preserved. An upstream 2xx response containing an error code is returned as HTTP 502. Missing or unparseable error fields use platform fallback values. Other top-level upstream fields, such as request_id, are not passed through in this error response.

This differs from generation failure after task creation: query responses report that through output.task_status, output.code, and output.message, as shown in the failed task example above.

Reference

For model parameters and media limits, see the official Bailian Wan 3.0 API reference. When integrating through Exchange Token, use the domain, platform API key, and task query flow on this page.